what ails
claude
Diagnose and govern the AI instruction systems — CLAUDE.md, AGENTS.md, .cursorrules, system prompts, hooks — that steer your coding agents.
npx @reporails/cli check
Sound familiar?
"My CLAUDE.md is too long and Claude ignores half the instructions"
"Copilot keeps ignoring my coding style preferences"
"I auto-generated my AGENTS.md and Codex treats it as noise"
"Cursor doesn't know how to run my tests or build the project"
"I switched from Copilot to Claude and had to rewrite everything"
"My .cursorrules references paths that no longer exist"
"I have no idea if my GEMINI.md is actually good or just vibes"
"My monorepo has one giant instruction file and agents get confused"
"I pasted my entire architecture doc into AGENTS.md and Codex chokes"
"My Copilot instructions are too vague to be useful"
"My instruction file has no markdown structure, just a wall of text"
"My Claude rules contradict each other across files"
Why this matters
Your AI coding agent reads a file. Quality of that file determines whether it follows your instructions or invents its own.
Please try to write good tests when possible.
Avoid using mocks if you can.vague — agent treats this as suggestion
## Testing
Do NOT use `unittest.mock`. Write integration tests against
a real `sqlite:///:memory:` database.specific — agent reliably follows
Built on research
Every rule is empirically validated
The rules are not opinions. Each one was derived from a controlled experiment on the 27,000-repo corpus and validated across Claude, Codex, Copilot, Gemini, and Cursor.
27k
repos analyzed
8.8M
diagnostic runs
5
LLMs cross-validated
92+
deterministic rules
See it in action
Run it. See what your instruction file is doing to your agent.
$ ails check
Reporails — Diagnostics
┌─ Main (2) 12 directive / 2 constraint · 69% prose
│ CLAUDE.md 8 dir / 1 con · 76% prose
│ ⚠ Missing tech stack declaration CORE:C:0034
│ 4 brief · 1 unbalanced
│
│ tests/CLAUDE.md 4 dir / 1 con · 29% prose
│ ⚠ Missing directory layout CORE:C:0035
│ 1 vague · 1 unbalanced
└─ 15 findings
┌─ Rules (8) 42 directive / 14 constraint · 48% prose
│ testing.md 6 dir / 3 con
│ ⚠ Instruction in heading CORE:S:0039
│ ... and 5 more (67 findings)
└─ 89 findings
── Summary ──────────────────────────────────────
Score: 7.0 / 10 ▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░░░░░░░░░ (0s)
Agent: Claude
Scope:
capabilities: 2 main, 8 rules, 4 skills
instructions: 54 directive / 112 prose (52%)
14 constraint
104 findings · 2 errors · 86 warnings · 16 info
Compliance: HIGH
Full diagnostics free for the first 100 registering users during beta
ails auth login
Beta phase
Run ails check anywhere — no account, no setup. Sign in with GitHub to see cross-file conflicts, reinforcement gaps, and per-instruction compliance scores.
npx @reporails/cli auth login